First Look at the Mantle Beneath an Active Back-Arc Basin

Contrib Mineral Petrol (2002) 143: 1–18 DOI 10.1007/s00410-001-0329-2

Yasuhiko Ohara Æ Robert J. Stern Æ Teruaki Ishii Hisayoshi Yurimoto Æ Toshitsugu Yamazaki Peridotites from the Mariana Trough: ﬁrst look at the mantle beneath an active back-arc basin

Received: 27 April 2001 / Accepted: 10 October 2001 / Published online: 11 December 2001 Springer-Verlag 2001

Abstract Two dives of the DSV Shinkai 6500 in the is no evidence of pervasive post-melting metasomatism Mariana Trough back-arc basin in the western Pacific or melt–mantle interaction. Spinel compositions plot in sampled back-arc basin mantle exposures. Reports of the field for abyssal peridotites. Clinopyroxenes show peridotite exposures in back-arc basin setting are very depletions in Ti, Zr, and REE that are intermediate limited and the lack of samples has hindered our between those documented for peridotites from the understanding of this important aspect of lithospheric Vulcan and Bouvet fracture zones (the American– evolution. The Mariana Trough is a slow-spreading Antarctic and Southwest Indian ridges, respectively). ridge, and ultramafic exposures with associated gabbro The open-system melting model indicates that the Ma- dykes or sills are located within a segment boundary. riana Trough peridotite compositions roughly corre- Petrological data suggest that the Mariana Trough spond to theoretical residual compositions after 7% peridotites are moderately depleted residues after partial near-fractional melting of a depleted MORB-type upper melting of the upper mantle. Although some peridotite mantle with only little melt or fluid/mantle interactions. samples are affected by small-scale metasomatism, there The low degree of melting is consistent with alow magma budget, resulting in ultramafic exposure. We infer that the mantle flow beneath the Mariana Trough Central Graben is episodic, resulting in varying magma Y. Ohara (&) supply rate at spreading segments. Ocean Research Laboratory, Hydrographic Department of Japan, Tokyo 104-0045, Japan E-mail: [email protected] Tel.: +81-3-35414387 Fax: +81-3-35413870 Introduction R.J. Stern Department of Geosciences, The Philippine Sea occupies a large part of the Western University of Texas at Dallas, Pacific, and is composed of several large back-arc basins. Richardson, TX 75083-0688, USA The Pacific Plate subducts beneath the Philippine Sea T. Ishii Plate along the Izu–Ogasawara–Mariana Trench Ocean Research Institute, (Fig. 1a). This region evolved through several episodes University of Tokyo, of arc formation, rifting and back-arc spreading (Karig Tokyo 164-8639, Japan 1971). The Kyushu-Palau Ridge and West Mariana H. Yurimoto Ridge are remnant arcs separated by the Parece Vela Department of Earth Sciences, Basin. The active arc is separated from the West Mari- Tokyo Institute of Technology, Tokyo 152-8551, Japan ana Ridge by an active back-arc basin, the Mariana Trough (Fig. 1a, b). The Mariana Trough is a slow- T. Yamazaki Institute for Marine Resources and Environment, spreading back-arc basin (Fryer 1995) that is propa- National Institute of Advanced Industrial Science gating northward through the Mariana arc system and Technology, Tsukuba 305-8567, Japan (Martinez et al. 1995). This is a back-arc basin where the Present address: Y. Ohara composition and evolution of basaltic lavas have been Department of Geology and Geophysics, studied extensively (Hart et al. 1972; Hawkins and Woods Hole Oceanographic Institution, Melchior 1985; Sinton and Fryer 1987; Volpe et al. 1987, Woods Hole, MA 02543-1539, USA 1990; Hawkins et al. 1990; Stern et al. 1990; Stolper and Editorial responsibility: T.L. Grove Newman 1994; Gribble et al. 1996, 1998). 2

Fig. 1. a Satellite altimetry map showing the tectonic feature of the Philippine Sea(datafrom Sandwell and Smith 1997). The rectangle shows the location of b. b Shaded structural image of the Mariana arc system illumi- nated from 90 (data from the predicted bathymetry of Smith and Sandwell 1997). Heavy dashed line indicates NVTZ (northern volcano tectonic zone) and SVTZ (southern volcano tectonic zone), and heavy plain line indicates spreading section of the Mariana Trough. Amagmatic rifting occurs in the Central Graben. The rectangle shows the location of Fig. 2

Reports of peridotite exposures in back-arc basins are limited to the Mariana Trough (Bloomer et al. Geologic background and sample location 1995; Stern et al. 1996, 1997) and the Parece Vela Basin (Ohara et al. 1996). Of these, only the Mariana Styles of extension vary along the strike in the Mariana Trough is presently active. One dredge haul during the Trough, which can be divided into four parts (Fig. 1b; 1991 Tunes 7 expedition aboard the R/V Thomas Martinez et al. 1995; Gribble et al. 1998): Washington first recovered back-arc basin mantle 1. The northern volcano-tectonic zone (NVTZ) – char- rocks (Stern et al. 1996). Two dives of the DSV acterized by asymmetric rifting accompanied by fis- Shinkai 6500 in 1996 mapped and sampled exposures sure eruption along normal faults and point source of back-arc basin mantle in the Mariana Trough volcanism. (Stern et al. 1997). Stern et al. (1996) argued by 2. The southern volcano-tectonic zone (SVTZ) – char- analogy with mid-ocean ridges (Mutter and Karson acterized by localization of volcanic and tectonic 1992; Tucholke et al. 1998) that slow back-arc exten- activity and southward-progressive separation of the sion may be accomplished amagmatically. Although extension axis from the arc. amagmatic tectonics may be common in certain phases 3. The Central Graben – deeps produced by mechanical of back-arc basin evolution based on the recent extension that expose back-arc basin lower crust and mapping study of the Philippine Sea (Ohara et al. upper mantle (Stern et al. 1996). 2001), little is known at present about the importance 4. South of 19.7N – extension occurs by slow seafloor of this process. spreading. The basin is widest at 18N, where full- In this paper, we describe in detail, for the first time, spreading rates are between 3 and 4.4 cm/year the petrology of upper mantle peridotite from the (Hussong and Uyeda 1981), corresponding to those Mariana Trough, and use petrographic, mineral chemi- of slow-spreading mid-oceanic ridges. cal, and ion microprobe trace element compositions to understand how these formed. In addition to the appli- These variations mark different stages in the evolu- cation of this information in understanding back-arc tion of the back-arc basin because of northward-pro- basin lithosphere evolution, these data fill a critical hole pagating rifting and spreading. Martinez et al. (1995) in our understanding of ophiolites. The lack of samples argued that true seafloor spreading exists only south of of back-arc basin mantle has limited our understanding 19.7N, although Yamazaki et al. (1993) argued that of ophiolite formation because back-arc basins are spreading occurs as far north as 22N based on magnetic thought to be very common in the inventory of data. Rifting occurs in the NVTZ, the SVTZ, and the ophiolites (Bloomer et al. 1995). Central Graben (Martinez et al. 1995; Baker et al. 1996). 3

Mariana Trough (5,700 m), and is remarkably deep even compared with the anomalously deep 3.6-km mean depth of the Mariana Trough back-arc basin spreading axis (Stern et al. 1996). A single dredge in the Southern Basin was performed during the 1991 Tunes 7 expedition aboard the R/V Thomas Washington. The dredge (Fig. 2; D45; 2002.5¢N, 14404.0¢E; 5,175–4,800 m depth) recovered a diverse suite of upper mantle and crustal rocks, includ- ing peridotites, gabbros, and felsic plutonic rocks. The petrologic data for the Central Graben crustal suite is clearly more similar to MORB or back-arc basin crust than it is to arc crust (Stern et al. 1996). The isotopic data for D45 crustal samples are also indistinguishable from basalts from portions of the Mariana Trough undergoing seafloor spreading and distinct from those of the Mariana arc. Stern et al. (1996) concluded that the dredge recovered samples from a talus pile at the base of the eastern scarp of the Central Graben, and argued that the Southern Basin exposes a back-arc basin crustal and upper mantle section, with the crustal suite remarkably heterogeneous with gabbros, basalts, and felsic plutonic Fig. 2. Bathymetry of the Southern Graben and track lines of DSV rocks. Shinkai 6500 dives 358 and 359. D45 represents the approximate The DSV Shinkai 6500 dives 358 and 359 were line of the Tunes 7 dredge 45 by R/V Thomas Washington (Stern conducted in 1996 in the Southern Basin in order to et al. 1996). Contours in 100-m intervals understand this unique upper mantle section (Fig. 2). Dive 358 recovered 14 samples from five sites, in- In the Central Graben, individual grabens are located cluding gabbros, basalts, tonalite, and serpentinites. within a20–30-km-wide sinuous bathymetric low This dive is inferred to have sampled debris flows that between 21 and 19.7N (Fig. 1b; Stern et al. 1996). The originated from higher up the scarp (Fig. 3; Stern et al. Central Graben is well delineated in the free-air gravity 1997). In contrast, dive 359 traversed mountainous anomalies as a band of relative lows between about 0 outcrops and steep exposures. Twenty-four samples and 25 mgal. Free-air gravity anomalies from distinct recovered from six sites along the dive 359 track sub-lows within this band and reflect the bathymetric represent an upper mantle suite of 14 peridotites and sub-basins (Martinez et al. 1995). Individual grabens are 10 gabbros, inferred as younger dikes or sills (Table 1; separated by along-strike bathymetric highs. Seismic Fig. 3; Stern et al. 1997). There is no significant Mn lines crossing the grabens (Yamazaki et al. 1993) show coating on any of these samples, suggesting these that they are fault bounded, relatively sediment free, surfaces have not been exposed for very long. In ad- generally asymmetric, and lack evidence of active vol- dition, the presence of talus blocks observed during canism. However, recent geophysical mapping of the northern Mariana Trough suggests that a neovolcanic zone may exist in parts of the Central Graben (Yama- Fig. 3. Schematic bathymetric profile of the east wall of the zaki and Stern 1997). The overall structure of the Cen- Southern Basin, the Central Graben of the Mariana Trough, with tral Graben is a complex graben that is composed of inferred subsurface lithologies (modified after Stern et al. 1997). No four asymmetric deeps. The southernmost of these deeps vertical exaggeration. Approximate positions of peridotite and gabbro samples recovered during dive 359 are shown (S denotes the (between 2010¢Nand1950¢N) is adown-to-the east station number). Stern et al. (1997) inferred that the petrologic half-graben (Fig. 2; ‘‘Southern Basin’’; Stern et al. 1997). Moho (the shallowest occurrence of mantle peridotites) may be The Southern Basin contains the deepest point in the exposed at a water depth about 4 km 4 the DSV Shinkai 6500 dive 358 suggests that most Stern et al. (1997) argue that the oceanic Moho may material has been reworked during tectonic events. not mark a sharp boundary between peridotite below Dive 359 documented several narrow, sediment-cov- and mafic rocks above, as inferred from seismic refrac- ered benches, suggesting that faulting is accomplished tion studies, but that the underlying mantle peridotite by a series of parallel, down-dropped blocks that had contains abundant mafic intrusions. These intrusions been displaced along steeply west-dipping normal become more common as the Moho is approached; thus, faults (Stern et al. 1997). The eastern wall of the the Moho marks the shallowest occurrence of mantle Southern Basin exposes a cross section of back-arc peridotites (Fig. 3). Stern et al. (1997) thus inferred the basin crust and upper mantle, with mantle exposed at petrologic Moho on the eastern wall of the Southern depths of greater than 4,530 m along the track of dive Basin is exposed at a water depth of about 4 km, 359 (Table 1; Figs. 2 and 3; Stern et al. 1997). although additional DSV Shinkai 6500 dives are necessary to prove this.

Table 1. List of the studied samples from the DSV Shinkai 6500 Petrography dives 358 and 359

Dive Station Depth Sample Remarks Nineteen peridotite and gabbro samples taken from the (m) dives 358 and 359 of DSV Shinkai 6500 were used in this study. Primary modal compositions of the peridotites 358 1 5,603 358R1-1 Gabbro 2 5,560 No sample for this study were determined by point counting. Because the altera- 3 5,430 No sample for this study tion products form characteristic pseudomorphs after the 4 5,162 358R4-4 Serpentinite primary phases, it is possible to determine the primary 5 4,790 358R5-2 Serpentinite mineral assemblage even when no relict mineral remains. 359 1 5,200 359R1-1 Gabbro/peridotite It is nevertheless diﬃcult to obtain modal compositions boundary for some samples because of injections of veins or ex- 359R1-2 Peridotite tensive serpentinization. Representative primary modes 359R1-3 Gabbro 359R1-4 Gabbro of the peridotites along with the mean value from the 2 5,149 359R2-1 Peridotite American–Antarctic and Southwest Indian ridges peri- 359R2-2 Peridotite (impregnated) dotites (from the Vulcan and Bouvet fracture zones, 359R2-3 Peridotite respectively; Dick 1989) are listed in Table 2. 3 5,061 359R3-3 Peridotite 4 4,957 359R4-2 Peridotite (impregnated) 359R4-3 Peridotite Peridotites 359R4-4 Peridotite (impregnated) 5 4,793 359R5-3 Peridotite 359R5-5 Peridotite (impregnated) Serpentinized harzburgites dominate the peridotites; no 6 4,536 359R6-1 Peridotite dunites were recovered during dives 358 and 359. The 359R6-2 Peridotite rocks are highly serpentinized and have veins of hydrous 359R6-3 Peridotite silicates and rare carbonates. Secondary modal compo-

Table 2. Representative primary modal compositions of the 359R1-2 is performed on 1·1-mm grid interval. No correction for Mariana Trough peridotites. Ol Olivine; Opx orthopyroxene; Cpx diﬀerential volume expansion during serpentinization has been clinopyroxene; Sp spinel; Pl plagioclase. Each analysis is performed made. Mean Vulcan and Bouvet fracture zones data are from Dick on 0.67·0.60-mm grid interval except for 359R1-2. Analysis for (1989) Sample Ol Opx Cpx Sp Pl Total Total points Lithology counted

(vol%) 359R1-2 72.91 24.15 1.13 1.81 0.00 100.00 443 Harzburgite 359R2-1 75.09 18.62 5.05 1.25 0.00 100.00 1,445 Lherzolite 359R2-3 86.83 11.96 0.12 1.09 0.00 100.00 1,739 Harzburgite 359R3-3 73.82 22.86 0.98 2.34 0.00 100.00 1,627 Harzburgite 359R4-3 75.44 22.31 0.69 1.56 0.00 100.00 1,600 Harzburgite 359R6-1 85.10 12.49 1.90 0.51 0.00 100.00 1,369 Harzburgite 359R6-2 81.40 17.49 0.50 0.61 0.00 100.00 1,812 Harzburgite 359R6-3 76.01 21.52 1.29 1.18 0.00 100.00 1,701 Harzburgite Mariana Trough mean 78.32 18.93 1.46 1.29 0.00 100.00 1,467 Harzburgite 1r 5.37 4.68 1.55 0.60 0.00 0.00 439 Vulcan FZ mean 71.13 20.56 7.33 0.96 0.03 100.01 – Lherzolite 1r 3.78 2.64 2.42 0.35 0.04 0.03 – Bouvet FZ mean 81.55 16.05 1.58 0.48 0.18 99.84 – Harzburgite 1r 4.18 3.44 1.31 0.29 0.40 4.81 – 5 sitions were given by Stern et al. (1997), and the degree were 15 kV accelerating voltage and 12 nA specimen of serpentinization or alteration is approximately 40 to current. A 1 lm beam size was used for all minerals 95%. The peridotites consist of serpentinized olivine and except pyroxenes, which were measured using a broader bastitized orthopyroxenes (10 mm) with minor clino- beam size (10 lm) to average the compositional segre- pyroxene (2 mm; Fig. 4a); however, some original gation caused by exsolution. Correction procedures are phases are generally preserved. The primary mineral those of Bence and Albee (1968). Total Fe is assumed assemblages are weakly deformed (Fig. 4b) and the equal to Fe2+ except for spinel. Fe3+ content of spinel freshest portions of the rocks exhibit porphyloclastic was calculated based on stoichiometry. The mineral textures (Fig. 4c). No peridotite mylonites were recov- compositions presented in this paper are averages of ered from the sample suite in the Southern Basin, multiple core analyses (generally 10–20 analyses). although peridotite mylonites are major constituents at slow-spreading ridges from the Atlantic and Indian Oceans (Jaroslow et al. 1996). Spinel is light reddish- Olivine brown and generally occurs as highly irregular shape intergrown with orthopyroxene (Fig. 4d), or with a Olivine is the major component of all peridotites. The symplectitic habit (Fig. 4e). olivine data are presented in Table 3. The forsterite Four samples (359R2-2, 359R4-2, 359R4-4, and content is relatively constant in each sample 359R5-5) show textural evidence of later melt or fluid (Fo=90.5) except for impregnated samples, which impregnation by veins, although we cannot rule out the show alower forsterite content (Fo=83–85). The NiO possibility that these veins are caused by alteration. content of primary olivines ranges from 0.44 to The inferred impregnated parts contain altered plagio- 0.32 wt%, typical of mantle peridotites. All but the clase and euhedral to subhedral rounded spinel as diffuse impregnated two samples (359R2-2 and 359R4-4) plot veins, whereas irregular-shaped spinels are found in the within the olivine–spinel mantle array (OSMA; Fig. 5; harzburgitic host of the rock. Veins clearly cut the ori- Arai 1994). Arai (1994) argues that OSMA is a residual ginal porphyroclastic texture in 359R2-2 (Fig. 4f), host- peridotite array and that cumulates plot off this trend to ing secondary amphibole rimming clinopyroxene. These the right. petrological relationships suggest that the veins intruded The gabbros show lower Fo content and have a the host peridotite after initial melting and deformation, larger range (Fo=59–75; Table 3) which is suggestive of and that these intruded as low temperature magmatic crystal fractionation, or of slow cooling with partial veins, or subsolidus, secondary, metasomatic veins. re-equilibration between crystal and interstitial residual liquids. Gabbros Spinel Nearly half of the samples recovered during dive 359 are gabbros that are inferred to be dykes or sills intruded Nearly all of the peridotite samples analyzed contain into the mantle peridotites (Figs. 3 and 4g). These are some unaltered spinel. The spinel data are presented in remarkably fresh and evolved, with abundant clinopy- Table 4. A plot of Mg/(Mg+Fe) ratio (=Mg#) vs. Cr/ roxene (5 mm) and little olivine (2 mm; Stern et al. (Cr+Al) (=Cr#) ratio in spinel is shown in Fig. 6a. 1996). Another remarkable feature of Central Graben Primary spinel shows virtually constant Cr# ranging gabbros is abundant red-brown hornblende (2 mm), from 0.24 to 0.29, with Mg# ranging from 0.69 to 0.73. which occurs both as rims on clinopyroxene and as The Cr# of spinel in abyssal peridotites is a good indi- primary phases (Fig. 4h). This is consistent with the cator of the degree of melting (Dick and Bullen 1984), much greater water content of back-arc basin basaltic with the low Cr# spinels representing less-depleted per- magmas. Mariana Trough basaltic glasses contain up to idotite and high Cr# spinels representing more-depleted 2.8 wt% H2O (Gribble et al. 1996), much greater than peridotite. TiO2 content of primary spinel is virtually the <0.2 wt% H2O typical of MORB (Michael 1988). zero, ranging from 0 to 0.05 wt% (Fig. 6b), which The Mariana Trough gabbros have coexisting clinopy- markedly contrasts the plagioclase-bearing abyssal per- roxene and plagioclase that are indistinguishable from idotites from the equatorial Atlantic (Dick and Bullen mid-ocean ridge gabbros, and thus are best interpreted 1984; Bonatti et al. 1993; Seyler and Bonatti 1997). as forming in association with the Mariana Trough Spinels in plagioclase-bearing peridotites are richer in Ti spreading (Stern et al. 1996). and Fe3+ and tend to be more Cr-rich compared with plagioclase-free peridotites from the same locality (Dick and Bullen 1984; Seyler and Bonatti 1997). Impregnated Mineral chemistry samples with Ti-rich spinel (359R2-2, 359R4-4, and 359R5-5; TiO2 content are 0.092 to 0.212 wt%) also Mineral compositions were analyzed with a JEOL have Ti-rich clinopyroxene. These spinels also have JCXA-733 electron microprobe at the Ocean Research higher Cr# (0.26–0.36) and lower Mg# (0.61–0.71). A Institute, University of Tokyo. The operating conditions positive correlation between Ti contents in spinel and 6

Fig. 4. a–h Microphotographs of thin sections and enlarged photos of thin sections, all taken with transmitted light. a Common habit of orthopyroxene and clinopyroxenes in the Mariana Trough peridotites. Sample 359R2-1, crossed polarizers. b Weakly deformed orthopyroxene in sample 359R6-3, crossed polarizers. c Serpentinized harzburgite with porphyloclastic texture in sample 359R6-3, crossed polarizers. d Irregular spinel intergrown with bastite after orthopyroxene in sample 359R2-3, plane-polarized light. e Symplectic spinel in sample 359R2-3, plane-polarized light. f Veined harzburgite. Sample 359R2-2, crossed polarizers. The vein (indicated by white arrow) clearly cuts the original porphyroclastic texture. g Con- tact relationships between serpentinized peridotite and gabbro. White arrows indicate contact boundary. Sample 359R1-1, crossed polarizers. h Red-brown hornblende in gabbro. Sample 359R1-3, plane-polarized light. Ol Oli- vine; OPx orthopyroxene; Cpx clinopyroxene; Sp spinel; Pl plagioclase; Hb hornblende

clinopyroxene is known in the abyssal peridotites (Dick Figure 8a–d shows Mg# vs. Al2O3,Cr2O3, TiO2, and Bullen 1984; Seyler and Bonatti 1997). and Na2O in clinopyroxene, respectively. The Mg# of primary clinopyroxene ranges from 0.91 to 0.92, and the Al2O3 content varies from 4.9 to 5.5 wt%. Pyroxenes The Cr2O3, TiO2,andNa2O contents of primary clinopyroxene are variable; Cr2O3 content ranges from Orthopyroxene and clinopyroxene compositions 1.0 to 1.4 wt%, TiO2 content from 0.09 to 0.17 wt%, (Tables 5 and 6) are approximately constant, mostly and Na2O content from 0.05 to 0.27 wt%. Al2O3 falling in the enstatite and the diopside fields, respec- and Cr2O3 levels of Mariana Trough peridotites lie tively. Figure 7a–c shows Mg# vs. Al2O3,Cr2O3,and approximately midway between the least and most TiO2 in orthopyroxene, respectively. The Mg# of pri- depleted abyssal peridotites. Na2O and TiO2 contents mary orthopyroxene ranges from 0.90 to 0.91. The in Mariana Trough peridotites are near the Al2O3 content of primary orthopyroxene ranges from most depleted compositions of abyssal peridotites 3.7 to 4.7 wt%. The orthopyroxene of impregnated overall. The clinopyroxene of impregnated samples sample 359R2-2 and 359R4-4 plots off the main (359R2-2, 359R4-4, and 359R5-5) plot off the main cluster defined by the Atlantic and Indian Ocean cluster defined by the Atlantic and Indian Ocean peridotites. peridotites. 7

Table 3. EPMA analyses of olivine in the Mariana Trough peridotites and gabbros (wt%). No. anal. Number of analyses making up the average composition. Fo forsterite content. Olivine in 359R4-4 and 359R1-1 retains two diﬀerent disequilibrium compositions (denoted as a and b)

Sample Peridotite Gabbro No. anal. 359R1-2 359R2-1 359R2-2 359R2-3 359R3-3 359R4-3 359R4-4a359R4-4b 359R6-3 359R1-1a 359R1-1b 26 10 16 34 17 41 14 4 47 15 20

SiO2 39.99 40.55 38.76 40.74 40.31 40.43 39.76 38.36 40.80 37.48 35.38 TiO2 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Al2O3 0.08 0.01 0.02 0.02 0.02 0.02 0.01 0.01 0.02 0.04 0.04 FeO 9.28 9.11 15.75 9.07 9.52 9.09 14.66 21.83 9.17 22.79 34.98 MnO 0.13 0.15 0.26 0.17 0.13 0.17 0.23 0.41 0.17 0.41 0.74 MgO 49.96 49.96 44.23 50.04 49.49 49.81 45.44 39.78 50.05 38.20 27.93 CaO 0.04 0.02 0.04 0.04 0.03 0.03 0.03 0.03 0.02 0.05 0.09 NiO 0.35 0.41 0.23 0.42 0.32 0.44 0.36 0.20 0.41 0.22 0.08 Total 99.83 100.22 99.28 100.49 99.81 100.00 100.50 100.62 100.65 99.18 99.24 Fo 90.6 90.7 83.4 90.8 90.3 90.7 84.7 76.5 90.7 74.9 58.7 1 (Fo) 0.2 0.2 0.3 0.1 0.2 0.2 0.3 0.2 0.2 0.3 0.2 Remarks Impregnated Impregnated Impregnated

The Mg# of gabbroic orthopyroxene ranges from 0.69 to 0.70, and the Al2O3 content from 1.1 to 1.2 wt%. The Mg# of gabbroic clinopyroxene ranges from 0.71 to 0.75, and the Al2O3 content from 1.8 to 2.3 wt%. These data again show the distinctly fractionated nature of the gabbroic pyroxenes.

Plagioclase

Plagioclase in the impregnated peridotite samples is totally altered and thus we cannot know the original compositions. Compositions of plagioclase in the gabbros are presented in Table 7. These plagioclase have a constant anorthite (An=49–54), which, again, is quite fractionated.

REE and trace elements in clinopyroxenes Fig. 5. Fo content of olivine vs. Cr# of coexisting spinel in Mariana Trough peridotites. Impregnated samples are shown by filled Selected peridotite clinopyroxenes were analyzed in-situ diamonds. The region between the broken lines is the olivine–spinel for REEs, Zr, and Ti using a Cameca IMS-3f ion mantle array (OSMA), where mantle-derived spinel peridotites plot microprobe both at the Tokyo Institute of Technology (Arai 1994). Note that impregnated samples, 359R2-2 and 359R4-4 and the Woods Hole Oceanographic Institution. Ana- (olivine composition is 359R4-4a; Table 3), plot off the OSMA. lytical procedures at the Tokyo Institute of Technology Olivine is absent in the other impregnated sample (359R5-5) are described by Yurimoto et al. (1989), and Wan and Yurimoto (1993), and at the Woods Hole Oceano- impregnated sample 359R2-2. These LREE-depleted graphic Institution are described by Shimizu et al. (1978) REE patterns are similar to those reported by Johnson and Johnson et al. (1990). Four clinopyroxenes from et al. (1990), Sobolev and Shimizu (1992, 1993) and Ross peridotite tectonites taken at stations 1, 2, 4, and 6 were and Elthon (1997) for clinopyroxenes from abyssal per- analyzed, along with two clinopyroxene from impreg- idotites. Clinopyroxene from impregnated sample nated samples collected at stations 2 and 4 (Fig. 3 and 359R2-2 exhibits aflatREE patternwith anegative Eu Table 8). Clinopyroxenes in the impregnated samples anomaly, suggesting that the equilibrium melt experi- 359R2-2 and 359R4-4 lie within the host peridotite, not enced significant plagioclase fractionation. 359R4-4 within the veins. exhibits LREE inflection in the REE pattern. This U- Chondrite-normalized REE patterns for clinopyrox- shaped REE pattern is generally attributed to early stage enes are shown in Fig. 9a. Clinopyroxenes exhibit LREE- of metasomatism. Zr and Ti concentrations in clino- depleted REE patterns with the notable exception of the pyroxenes are plotted in Fig. 9b. It is evident that the 8

Fig. 6. a,b Spinel compositional plots, fields for abyssal peridotite are from Dick and Bullen (1984). a Mg# vs. Cr#. The y axis length is equal to two times the x axis, reflecting the relative molecular proportions of R2+ and R3+ cations. Impregnated samples are shown by filled diamonds. b TiO2 wt% vs. Cr# in spinel. Hatched area is the range for most abyssal plagioclase-bearing peridotites (Dick and Bullen 1984). Note that impregnated samples, 359R4-4 and 359R5-5, have higher TiO2 content

impregnated sample 359R2-2 is enriched in both Zr and Ti compared with the other peridotite clinopyroxenes. Employing the models of Johnson et al. (1990), clinopyroxenes in the Mariana Trough peridotites (excluding 359R2-2 and 359R4-4) have residual com- Number of analyses making up the average composition positions that can be explained by 5% near-fractional melting of adepleted, MORB-type upper mantle. These observations agree with other petrological data, which No. anal. indicate that Mariana Trough peridotites are moderately depleted.

Discussion

Petrological characteristics of Mariana Trough peridotites

Partial melting of the upper mantle and extraction of basaltic melts produces ultramaﬁc tectonites as solid residues. Experimental studies indicate that progressive melting of spinel lherzolite at 10–20 kb rapidly eliminates clinopyroxene and gradually reduces the proportion of orthopyroxene (Mysen and Kushiro 1977; Jaques and Green 1980). As melting proceeds, forsterite and NiO contents of olivine, Mg# of pyroxenes, and Cr# of spinel increase, and Al2O3 content of spinel and pyroxenes and of the whole rock decrease. The slightly deformed porphyroclastic textures of relict primary minerals, the 0.022.19 0.05 0.09 2.35 2.52 0.00 2.07 0.05 0.05 1.42 0.75 0.00 1.04 0.21 2.40 0.00 0.13 2.43 3.70 0.01 2.26 0.01 2.37 0.00 2.16 45.64 46.09 43.7923.09 46.30 22.02 22.91 46.17 42.47 21.81 46.30 22.77 26.78 39.44 22.43 43.19 36.40 27.23 42.21 24.73 29.91 43.25 45.53 26.04 25.06 22.68 192221510181421152010640

Peridotite 359R1-2 359R2-1 359R2-2 359R2-3 359R3-3 359R4-2 359R4-3 359R4-4 359R5-3 359R5-5 359R6-1 359R6-2 359R6-3 irregular shape of spinels, and the mineral modes of Mariana Trough peridotites further indicate an origin as

EPMA analyses of spinel in the Mariana Trough peridotites (wt%). residual mantle that has undergone high-temperature deformation under upper mantle plastic ﬂow conditions. 3 3 3 2 O O O (Mg#)(Cr#) 0.005 0.007 0.001 0.002 0.006 0.004 0.009 0.003 0.011 0.003 0.007 0.004 0.013 0.011 0.011 0.028 0.043 0.003 0.018 0.021 0.005 0.003 0.006 0.005 0.007 0.007

2 2 The chemistry of Mariana Trough peridotites is also 2 r r FeOFe 11.77NiOTotal 13.03Mg#1 13.68Cr# 0.25 101.081 Remarks 11.93 101.23 0.732 0.26 0.253 99.73 12.13 0.704 0.23 12.71 0.243 100.26 0.683 0.29 100.67 0.260 11.73 0.725 99.97 0.23 0.240 16.84 0.725 Impregnated 99.65 0.15 0.249 0.705 13.27 100.76 0.297 13.94 0.25 0.729 100.73 0.245 12.31 0.606 0.25 99.98 Impregnated 0.317 13.57 0.691 100.35 0.25 0.278 0.667 12.47 101.19 Impregnated 0.26 0.356 100.65 0.712 0.25 0.293 Impregnated 0.685 0.25 0.280 0.713 0.250 0.26 Al MnOMgOCaOCr 0.12 17.99 0.00 0.16 17.27 0.00 16.35 0.14 0.00 17.69 0.17 17.76 0.00 0.14 16.92 0.00 0.13 17.74 0.01 0.17 14.19 0.00 0.21 16.63 0.00 15.43 0.23 0.01 17.07 0.20 0.01 16.54 0.19 17.40 0.00 0.15 0.00 0.14 0.01 Table 4. Sample No. anal. TiO compatible with a residual origin: their high contents of 9 Enstatite En Gabbro ) b and a Number of analyses making up the average composition. No. anal.

Fig. 7. Orthopyroxene compositional plots for Mariana Trough peridotites and gabbros, with ﬁelds for Atlantic and Indian Ocean peridotites for comparison (Johnson et al. 1990; Bonatti et al. 1992; Cannat et al. 1992; Johnson and Dick 1992; Cannat and Seyler

wollastonite content. Orthopyroxene in 359R4-4 retains two different disequilibrium compositions (denoted as 1995; Ross and Elthon 1997). a Mg# vs. Al2O3 wt%. Impregnated samples 359R2-2 and 359R4-4 plot off the main cluster defined by

Wo Atlantic and Indian Ocean peridotites. Orthopyroxene is absent in two other impregnated samples, 359R4-2 and 359R5-5. b Mg# vs. Cr2O3 wt%. c Mg# vs. TiO2 wt%

compatible elements such as Ni, Cr, and Mg, and their

0.044.45 0.02 4.20 0.030.86 3.86 0.02 0.83 4.26 0.04 0.80 4.65 0.01 0.83 4.53low 0.08 0.87 contents 4.37 0.07 0.93 3.49 of incompatible 0.89 0.02 4.68 0.00 0.74 4.20 elements 0.05 1.02 4.00 0.02 such 0.94 3.73 0.39as 0.98 1.20 Ti and 0.32 0.70 1.08 Na. 0.01 0.00 53.76 55.18 52.93 54.65 54.01 54.58 54.27 54.63 54.12 54.84 53.98 55.01 52.10 53.00 359R1-228 359R2-1 359R2-2 359R2-3 9 359R3-3 359R4-3 359R4-4a359R4-4b 14 359R5-3 359R6-1 359R6-2 15 359R6-3 358R1-1 359R1-4 29 16 19 10 19 10 7 42 30 49 Peridotite

ferrosilite content; The Marina Trough peridotites are LREE-depleted EPMA analyses of orthopyroxene in the Mariana Trough peridotites and gabbros (wt%).

Fs with low incompatible element abundances and have spinels with relatively low Cr# (0.25). The moderately 3 3 2 O 0.01 0.01 0.08 0.00 0.01 0.00 0.04 0.03 0.03 0.01 0.01 0.00 0.04 0.04 2 O O (Mg#) 0.002 0.002 0.009 0.001 0.002 0.002 0.008 0.008 0.004 0.003 0.003 0.003 0.004 0.006 2

2 depleted (in terms of major elements) nature of the 2 r Al FeOMnOMgOCaONa 5.97 0.13 32.22NiO 1.89Total 5.88 0.21Mg# 32.851 0.04 8.83 29.79 1.47 99.37En 0.19Fs 0.906Wo 2.35 100.80 32.60 0.14 5.94Remarks 0.17 87.2 98.88 0.909 31.96 0.03 1.31 6.14 9.1 0.13 3.7 0.858 99.91 88.3 32.49 1.74 0.12 5.87 0.907 8.9 0.17 99.59 81.7 2.8 31.91 0.05 1.42 0.903 6.84 13.6 100.12 0.16 4.7 88.4 30.95 Impregnated 100.22 1.59 0.908 0.12 8.85 9.0 0.20 31.49 87.2 2.6 100.65 0.893 0.09 1.60 5.71 32.90 9.4 0.17 88.3 99.60 3.4 0.862 32.15 0.09 2.23 6.00 0.16 99.97 86.5 9.0 0.908 32.51 2.8 0.79 0.13 5.80 99.04 0.09 0.907 10.4 23.76 83.5 99.83 1.88 3.1 0.13 5.81 Impregnated 0.17 0.908 Impregnated 24.33 86.7 13.4 99.02 19.50 1.76 0.10 0.909 0.67 3.1 99.78 89.3 18.96 0.685 8.8 1.34 0.12 0.66 4.4 87.5 0.696 1.36 0.02 9.1 87.8 1.5 0.02 8.9 66.6 3.7 8.8 67.7 3.4 30.7 2.7 29.6 2.7 Cr TiO Table 5. content; SiO Sample No. anal. Mariana Trough peridotite is attested to by the modal 10

Table 6. EPMA analyses of clinopyroxene in the Mariana Trough peridotites and gabbros (wt%). No. anal. Number of analyses making 359R5-5, and 359R1-1 retains several diﬀerent disequilibrium compositions (denoted as a, b,andc)

Sample Peridotite No. anal. 359R1-2 359R2-1 359R2-2a359R2-2b 359R2-3 359R4-3 359R4-4a359R4-4b 15 12 28 11 6 6 15 9

SiO2 50.18 51.59 50.99 51.53 51.22 51.05 53.37 51.83 TiO2 0.16 0.19 0.18 0.91 0.12 0.17 0.16 0.31 Al2O3 5.35 5.10 3.96 2.52 4.86 5.53 3.03 3.85 FeO 2.64 2.63 3.36 5.00 2.65 2.44 3.35 4.80 MnO 0.08 0.09 0.12 0.19 0.11 0.11 0.12 0.19 MgO 16.35 16.28 16.13 16.83 16.48 16.07 16.72 15.39 CaO 23.00 23.29 22.39 21.18 23.06 23.39 23.32 22.39 Na2O 0.08 0.27 0.47 0.43 0.11 0.20 0.36 0.50 Cr2O3 1.21 1.29 1.22 0.14 1.03 1.37 0.96 1.16 NiO 0.01 0.05 0.01 0.01 0.06 0.09 0.05 0.03 Total 99.07 100.78 98.83 98.74 99.70 100.40 101.45 100.45 Mg # 0.917 0.917 0.896 0.858 0.917 0.922 0.899 0.851 1r (Mg#) 0.002 0.005 0.016 0.013 0.002 0.002 0.006 0.010 En 47.6 47.2 47.3 48.3 47.7 46.9 47.3 45.0 Fs 4.3 4.3 5.5 8.0 4.3 4.0 5.3 7.9 Wo 48.1 48.5 47.2 43.7 48.0 49.1 47.4 47.1 Remarks Impregnated Impregnated Impregnated Impregnated compositions and the chemistry of their relict primary introduce Fe, Ti, LREE, and other trace elements into minerals (Mg #, Al2O3 and TiO2 contents of olivine, the peridotite (Bonatti et al. 1993). This metasomatic spinel, and pyroxenes). Modal compositions of Mariana process could account for the ‘‘enriched’’ chemistry of Trough peridotites are intermediate between well-char- spinel and pyroxenes in impregnated samples. acterized peridotites recovered from fracture zones on The overall petrological characteristics of Mariana the American–Antarctic and Southwest Indian ridges Trough peridotites are very similar to those of perido- (Vulcan and Bouvet fracture zones, respectively; Dick tites drilled at the hole 920D, Ocean Drilling Program 1989; Table 2). Among the peridotites studied by (ODP) leg 153, at the MARK area of the Mid-Atlantic Johnson et al. (1990), those from the Vulcan Fracture Ridge, where magma channeling occurred in cold Zone are the least depleted and those from the Bouvet peridotites at a shallow upper mantle (Niida 1997). Fracture Zone are the most depleted. Compositions of spinels, orthopyroxene, and clinopyroxene in Mariana Trough peridotites consistently plot in the middle of the Melting model abyssal peridotite range (Figs. 6, 7, and 8), suggesting that the Mariana Trough peridotites are ‘‘moderately Models for magma genesis beneath mid-ocean ridges depleted’’ in aglobalsense. (Gast 1968), as well as recent melting experiments and Plagioclase is sometimes common in the abyssal observation of trace element abundances in clinopy- peridotite samples as veins or interstitial patches (Dick roxenes from abyssal peridotites (Johnson et al. 1990), and Bullen 1984; Bonatti et al. 1992; Arai and Mat- suggest that small volume melt fractions can be effi- sukage 1996; Seyler and Bonatti 1997). These plagio- ciently extracted from the melting mantle. Residues of clase-bearing peridotites are characterized by higher Mg/ such fractional melting display extremely low abun- Fe ratio and enrichment in incompatible trace elements dances of incompatible trace elements and extreme such as Ti (Dick and Bullen 1984; Seyler and Bonatti fractionation between those elements. Using ion 1997), and are generally considered to be products of microprobe analyses of clinopyroxene from abyssal melt–mantle interaction (Dick and Bullen 1984; Arai peridotites from the slow-spreading ridges in the and Matsukage 1996; Seyler and Bonatti 1997). No Atlantic and Indian Oceans, Johnson et al. (1990) have plagioclase-bearing peridotites are observed in the shown that equilibrium melting cannot produce Mariana Trough tectonite suite, although four impreg- observed trace element variations in clinopyroxene. nated samples from three dive stations were found There now is consensus that equilibrium melting of the (359R2-2, 359R4-2, 359R4-4, and 359R5-5). The oc- mantle will not sufficiently deplete Zr, Ti, and REE in currence of impregnated peridotites that are associated abyssal peridotite clinopyroxenes (Figs. 10 and 11) and with normal harzburgite and lherzolite may be caused by also not produce liquids with Na and trace element fractionated mafic melts intruding into peridotites that abundances comparable with primitive MORB glasses surround small and transient magma chambers beneath (Johnson et al. 1990). the Mariana Trough. The injection of fluids or melts as REE abundances in Mariana Trough peridotite clin- narrow veins (1–10 cm) in the shallow mantle could opyroxenes are compared with those in peridotites from 11

up the average composition. En enstatite content; Fs ferrosilite content; Wo wollastonite content. Clinopyroxene in 359R2-2, 359R4-4,

Gabbro

359R4-4c 359R5-5a359R5-5b 359R6-3 358R1-1 359R1-1a359R1-1b 359R1-1c 359R1-3 359R1-4 10 21 2 28 10 8 20 9 79 46

52.17 51.51 53.89 51.38 51.03 52.01 51.45 51.69 51.89 51.96 0.88 0.12 0.02 0.09 0.66 0.57 0.57 0.62 0.53 0.53 2.53 4.59 0.99 5.33 2.27 1.95 2.03 1.98 1.95 1.84 6.43 2.77 7.43 3.15 9.85 8.45 9.26 10.30 9.02 8.93 0.27 0.11 0.39 0.14 0.38 0.25 0.33 0.33 0.37 0.37 15.90 16.11 14.00 18.53 14.41 14.54 15.03 14.12 14.54 14.72 21.10 23.12 22.80 20.01 19.81 21.25 20.16 20.03 20.88 20.72 0.44 0.16 0.60 0.05 0.46 0.39 0.38 0.39 0.44 0.43 0.19 1.46 0.73 1.21 0.01 0.07 0.00 0.01 0.00 0.00 0.03 0.06 0.02 0.09 0.03 0.01 0.01 0.02 0.02 0.03 99.94 100.00 100.87 99.97 98.89 99.49 99.22 99.48 99.65 99.53 0.815 0.912 0.770 0.913 0.723 0.754 0.743 0.710 0.742 0.746 0.012 0.011 0.015 0.005 0.012 0.006 0.010 0.014 0.012 0.011 45.9 47.0 40.5 53.4 42.2 42.1 43.3 41.2 42.0 42.5 10.4 4.5 12.1 5.1 16.2 13.7 15.0 16.8 14.6 14.5 43.7 48.5 47.4 41.5 41.7 44.2 41.7 42.0 43.4 43.0 Impregnated Impregnated Impregnated the American–Antarctic and Southwest Indian ridges (Sobolev and Shimizu 1992, 1993). The retained melt (from the Vulcan and Bouvet fracture zones, respec- efficiently buffers highly incompatible elements from tively; Johnson et al. 1990; Figs. 9a, b). The Mariana extreme fractionation (Sobolev and Shimizu 1992, 1993). Trough peridotite clinopyroxenes show light REE de- The critical melting model with 1 wt% of retained melt pletions similar to Bouvet, but middle and heavy REE at 10% melting is grossly consistent with the clinopy- concentrations comparable to Vulcan (Fig. 9a). The ex- roxene data, although it disagrees somewhat for light tremely depleted nature of Bouvet peridotites were at- REEs and heavy REEs concentrations (Fig. 10; tributed to extensive melting associated with plume Appendix A). activity at the Bouvet hot spot (Johnson et al. 1990). In order to better model the melting process in the Estimating the degree of melting experienced by Mariana Trough, we calculated the REE concentrations Mariana Trough peridotites depends on assumptions of in Mariana Trough peridotite clinopyroxene using the the mantle composition when melting begins, the melt- open-system melting model developed by Ozawa and ing mode, and partition coefficients. Based on Nd and Shimizu (1995). The open-system melting model was Hf isotopic compositions of MORB, we chose aLREE- first proposed by Dick et al. (1984) and refined by depleted lherzolite for the initial material (Table 8; Johnson and Dick (1992). A quantitative model was Sobolev and Shimizu 1992). Partition coefficients for Ti, developed by Ozawa and Shimizu (1995), in which Zr, and REE used in the modeling are listed in Table 9. coupling of melting and melt extraction under the influx The incongruent non-modal melting mode (Table 10) is of aLREE-enriched exotic melt or fluid is assumed.We taken from Kinzler and Grove (1992). These parameters assume that the influxed material is silicate melt in are based on anhydrous melting experiments and thus equilibrium with clinopyroxene in impregnated sample may not be appropriate for modeling melting of Mari- 359R2-2. This assumption may not be appropriate for ana Trough peridotites because melting in back-arc ba- inferring the melting process at depth because the sins is generally hydrous (Stolper and Newman 1994; influxed melt that affected sample 359R2-2 may be a Gribble et al. 1998). However, recent melting experi- shallow-level fractionate of a transient magma chamber. ments, using hydrous peridotite, indicate that partition However, we consider that this approach utilizes the coefficients do not differ significantly between anhydrous available data in the best way possible, and the results and hydrous conditions (Gaetani et al. 1997). On this place significant constraints on mantle melting beneath basis, we assume that the parameters are still valid for the Mariana Trough. melting of Mariana Trough peridotite. The observed This model reproduces the observed REE pattern of REE patterns of Mariana Trough clinopyroxenes can be Mariana Trough peridotites with 7% melting (F=0.07) roughly explained by 5% near-fractional melting of depleted mantle retaining 0.5 wt% of interstitial (Johnson et al. 1990) of depleted MORB-type upper melt (a=0.005) and a mass influx rate of 0.003 mantle (Fig. 10; Appendix A). (b=0.003) (Fig. 10; Appendix A). If infiltrating fluid is Critical (continuous) melting is also possible, where- LREE-enriched, then heavy REE concentrations are by small degrees of melting are followed by incomplete primarily controlled by the degree of melting, not by melt extraction, with some melt retained in the residue the fraction of melt retained or fluid influx (Ozawa 12

Fig. 8. a–d Clinopyroxene compositional plots for Mari- ana Trough peridotites and gabbros, with ﬁelds for Atlantic and Indian Ocean peridotites for comparison (Johnson et al. 1990; Bonatti et al. 1992; Can- nat et al. 1992; Johnson and Dick 1992; Cannat and Seyler 1995; Ross and Elthon 1997). Symbols as in Fig. 7. a Mg# vs. Al2O3 wt%. Impregnated samples, 359R2-2, 359R4-4, and 359R5-5, showing wide compositional variations within the sample, are indicated. Clinopy- roxene is absent in impregnated sample 359R4-2. b Mg# vs. Cr2O3 wt%. c Mg# vs. TiO2 wt%. d Mg# vs. Na2O wt%

Table 7. EPMA analyses of plagioclase in the Mariana Trough actions were insigniﬁcant for generating the Mariana gabbros (wt%). No. anal. Number of analyses making up the Trough peridotites. This conclusion contrasts with average composition. An Anorthite content; Ab albite content; large-scale pervasive melt–mantle interaction docu- Or orthoclase content mented for peridotites from the East Romanche Sample Gabbro Fracture Zone, equatorial Atlantic (Seyler and Bonatti No. anal. 1997) or the ‘‘refertilization’’ process proposed by 358R1-1 359R1-1 359R1-3 359R1-4 49 12 41 38 Elthon (1992). The clinopyroxene data show that melting in the

SiO2 55.15 54.58 55.78 55.88 upper mantle beneath the Mariana Trough produced TiO2 0.05 0.00 0.05 0.06 melts similar to the ultra-depleted melt (in terms of Al2O3 27.73 28.47 27.83 27.74 incompatible trace elements) of Sobolev and Shimizu FeO 0.26 0.13 0.23 0.24 MnO 0.05 0.00 0.05 0.04 (1992, 1993). Fractional melting of the mantle led to MgO 0.05 0.02 0.05 0.04 extreme depletion, producing late-stage melts that were CaO 10.50 11.29 10.43 10.36 strongly depleted in incompatible trace elements. Na2O 5.81 5.37 5.89 5.97 Depletions of high field strength elements are common K O 0.09 0.05 0.07 0.07 2 in the upper mantle clinopyroxenes from beneath mid- Cr2O3 0.01 0.00 0.01 0.01 NiO 0.03 0.01 0.03 0.02 ocean ridges (Johnson et al. 1990, 1995; Sobolev and Total 99.73 99.94 100.41 100.44 Shimizu 1992, 1993). It appears that the production of such strongly depleted melts (in terms of incompatible An 49.7 53.6 49.3 48.7 trace elements) is a routine feature of magma genera- Ab 49.8 46.1 50.4 50.8 Or 0.5 0.3 0.4 0.4 tion beneath mid-ocean ridges and back-arc basin 1r (An) 1.4 1.0 1.5 1.1 spreading centers. and Shimizu 1995). Assuming a MORB-type depleted Peridotite and basalt from the Mariana Trough mantle source and the parameters in Tables 9 and 10, 7% melting best reproduces the observed HREE con- Modeling of REE, Zr, and Ti concentrations in residual centrations. At 7% melting, the influxed mass fraction clinopyroxenes indicate that most abyssal peridotites are of the initial solid is 0.021%, which is very small. not in equilibrium with MORB (Johnson et al. 1990; Concentrations of Zr and Ti in the clinopyroxenes are Sobolev and Shimizu 1992, 1993; Ross and Elthon consistent with both the open-system (a=0.005 and 1997). Most MORBs are products of mixing between b=0.003) and the critical melting model (Fig. 11). This magmas, and the ultra-depleted characteristics of melts suggests that pervasive melt or fluid–mantle inter- are irreversibly lost in that process (Johnson et al. 1990; 13

Table 8. Ion microprobe analyses of clinopyroxene in the Mariana WHOI analysis at Woods Hole Oceanographic Institution. Initial Trough peridotites. No. anal. Number of analyses making up the bulk mantle composition from Sobolev and Shimizu (1992). average composition. Numbers in parentheses are standard devia- Chondrite composition from Anders and Ebihara (1982) tions (1r). TITECH Analysis at Tokyo Institute of Technology; Sample Peridotite Initial bulk Calculated Chondrite No. anal. mantle initial 359R1-2 359R2-1 359R2-2 359R4-3 359R4-4 359R6-3 clinopyroxene 13 1 13 1

(ppm) Ce 0.02 0.05 (0.01) 10.46 0.05 2.00 (0.24) 0.02 0.92 6.91 0.616 Pr 0.01 – 1.98 0.04 – 0.02 – 0.093 Nd 0.28 0.30 (0.02) 12.02 0.29 1.14 (0.22) 0.17 0.82 5.61 0.457 Sm 0.43 0.43 (0.02) 5.58 0.54 0.68 (0.05) 0.22 0.33 2.32 0.149 Eu 0.27 0.19 (0.04) 1.37 0.21 0.26 (0.01) 0.11 0.13 0.88 0.056 Gd 0.93 – 7.15 0.98 – 0.65 – 0.197 Tb 0.19 – 1.21 0.19 – 0.14 – 0.035 Dy 2.09 1.59 (0.09) 10.90 2.16 2.04 (0.12) 1.45 0.61 4.00 0.245 Ho 0.41 – 2.13 0.43 – 0.32 – 0.055 Er 1.69 1.02 (0.08) 6.76 1.56 1.43 (0.21) 1.20 0.40 1.89 0.160 Tm 0.17 – 0.92 0.23 – 0.16 – 0.025 Yb 1.26 0.96 (0.07) 6.57 1.40 1.34 (0.06) 1.06 0.40 2.36 0.159 Lu 0.20 – 0.81 0.20 – 0.13 – 0.024 No. anal. 1 3 1 1 3 1 (ppm) Ti 983 1,259 (32) 2,707 1,075 1,092 (110) 812 1,090 4,288 436 Zr 0.5 0.9 (0.1) 69.3 1.0 2.6 (0.6) 0.3 7.5 28.1 3.9 Remarks TITECH WHOI TITECH TITECH WHOI TITECH

Sobolev and Shimizu 1992). The residual mantle of and Gribble et al. (1996) argued that Mariana Trough Mariana Trough also is not in equilibrium with basalts basalt results from adding a water-rich ‘‘subduction from this area, nor with the intrusive gabbros. component’’ to an N-MORB mantle source. Gribble We calculated the accumulated melt composition et al. (1998) argued that Mariana Trough basalts are using the melting model tested above (equilibrium often – but not always – wetter than MORB. We infer melting, fractional melting, critical melting, and open- that the Central Graben peridotites are the residues of system melting), and compared the results with spatially low degrees of melting of adepleted MORB-type upper associated basalts from the Central Graben (Fig. 12; mantle under a relatively hydrous condition. Appendix B; Gribble et al. 1998). As for the open-system melting model, the trace element pattern of the accumulated melt is strongly controlled by mass influx rate Tectonic implications and less by the degree of melting (Ozawa and Shimizu 1995). Trace element concentrations in the calculated Ultramafic exposures on the seafloor are regarded as accumulated melts after the critical melting model manifesting a low magma budget and relatively thin (a=0.01 and F=0.1) and open-system melting crust (Cannat et al. 1995). Karson et al. (1987) suggested (a=0.005, b=0.003 and F=0.07) approximate the that mantle rocks were exposed in the median valley composition of the spatially associated basalts (Fig. 12), south of the Kane Fracture Zone of the Mid-Atlantic consistent with the idea that MORBs are the post-seg- Ridge during purely tectonic phases of spreading, regation accumulation of small increment melts pro- without any associated igneous activity. Peridotite or duced during mantle melting events (Johnson et al. 1990; gabbro exposures away from a fracture zone require Sobolev and Shimizu 1992, 1993). thinning and stretching of the crust by low angle and Basalts from along the extensional axis of the listric normal faulting and by block rotations (Tucholke northern Mariana Trough show progressive northward et al. 1998). Listric normal faulting was first suggested changes in chemical and isotopic compositions, from by Dick et al. (1981) to explain the wide range of deep typical back-arc basin basalts that formed by seafloor crustal rocks exposed at inside corners of the Kane spreading through increasingly arc-like basalts, until Fracture Zone. Exposures of mantle rocks away from volcanic rocks that are indistinguishable from arc vol- fracture zones are not known along fast spreading canic rocks are found in the NVTZ (Gribble et al. 1998). ridges. This is because higher magma production rates Although we have no supporting petrological evidence and more continuous magmatic activity produce oceanic such as the occurrence of primary hydrous minerals, the crust of more uniform and substantial thickness. Mariana Trough peridotites could be the residues of By analogy with slow spreading mid-ocean ridges, hydrous mantle melting. Stolper and Newman (1994) slow back-arc spreading should occasionally be amag- 14

Fig. 10. REE patterns of clinopyroxene for equilibrium melting, the fractional melting, the critical melting, and the open-system melting models (Appendix A). Concentrations in influxed melt for the open-system melting model are also shown (Appendix A). The Gd values are interpolated. The model initial material is depleted MORB-type mantle (Sobolev and Shimizu 1992; Table 8) and the mineral modal composition is given in Table 10. Partition coefficients are given in Table 9, and the melting mode is after Fig. 9. a,b Ion microprobe data for clinopyroxenes from Mariana Kinzler and Grove (1992; Table 10). The range for the primary Trough peridotites (Table 8). Data for clinopyroxenes from the Mariana Trough peridotite data (samples 359R1-2, 359R2-1, Vulcan and Bouvet fracture zones are shown for comparison 359R4-3, and 359R6-3) is hatched. The definition of the parameters (Johnson et al. 1990). a Chondrite-normalized REE patterns. is as follows (open-system melting; Ozawa and Shimizu 1995): F is Chondrite normalizing values are from Anders and Ebihara (1982). degree of melting, a is mass ratio of retained melt in solid residue The Gd values for the samples 359R2-1 and 359R4-4 are after the critical degree of melting is reached, and b is the mass interpolated. Impregnated samples, 359R2-2 and 359R4-4 are influx rate, which is influxed mass fraction (by a LREE-enriched indicated by dotted lines. REE patterns of samples 359R1-2, exotic melt or fluid) of the initial solid mass divided by degree of 359R2-1, 359R4-3, and 359R6-3 have steeply sloping LREE- melting. The open-system melting model gives the solution for depleted patterns, whereas impregnated sample 359R2-2 shows perfect fractional melting (Shaw 1970) by setting both a and b rather flat REE pattern with an Eu anomaly. 359R4-4 exhibits equal to zero, and the solution for critical melting (Sobolev and LREE inflection in the REE pattern. The Vulcan and Bouvet data Shimizu 1993) by setting b equal to zero. The open-system melting define the upper and lower limits of published REE concentrations model a=0.005, b=0.003, and 7% melting (F=0.007) shows good in clinopyroxene, respectively. b Zr and Ti concentrations agreement with the observed Mariana Trough peridotite data. The other possible model to reproduce the Mariana Trough data is critical melting (a=0.01 and 10% melting) matic. Recent bathymetric mapping of the Philippine Sea back-arc basins suggest that amagmatic tectonics is evolution of other back-arc basins such as the Sumisu a common phenomena in certain phases of back-arc Rift in the Izu-Ogasawara arc (Taylor et al. 1991). basin evolution (Ohara et al. 2001). Latest geophysical The low degree of melting (7%) estimated for the survey of the northern Mariana Trough revealed that Mariana Trough peridotites is consistent with a low the structure of the Central Graben is essentially the magma budget, resulting in ultramafic exposure in the same as those of slow-spreading mid-oceanic ridges Central Graben. Intrusive relationships between frac- (Yamazaki et al. 1999). Yamazaki et al. (1999) argue tionated gabbro and mantle peridotite also may attest to that the Central Graben is located within a segment a magma-starved spreading environment (Cannat et al. boundary. The location of down-to-the-east half graben 1992). Ar–Ar dating on the hornblende of crustal along the eastern margin of the Central Graben is con- gabbro-diorite dredged during the 1991 Tunes 7 expe- sistent with the presence of awest-dipping, low-angle dition yields 1.8±0.6 Ma (Stern et al. 1996). Based on normal fault (Stern et al. 1996, 1997). Similar structures the age data, Stern et al. (1996) argued that magmatic have been documented for the earliest stages in the activity was important before 1.8 Ma, but that 15

Fig. 12. REE patterns in accumulated melt for the equilibrium melting, the fractional melting, the critical melting, and the open- system melting models (Appendix A). REE patterns for Mariana Trough basalts from the Central Graben (hatched area; Gribble et al. 1998) are shown for comparison. Concentrations in inﬂuxed melt for the open-system melting model are also shown. The Gd values are interpolated. REE patterns in the accumulated melts using the critical melting model (Sobolev and Shimizu 1992) and open-system melting (Ozawa and Shimizu 1995) approximate the composition of the spatially associated basalts

magma supply rates beneath spreading segments. Our Fig. 11. Zr and Ti concentrations in modeled clinopyroxenes (Appendix B). The primary Mariana Trough peridotite data preferred model for the evolution of the Central Graben (samples 359R1-2, 359R2-1, 359R4-3, and 359R6-3) are also starts with magmatic rifting, associated with relatively plotted. The open-system melting model (a=0.005 and b=0.003) ‘‘wet’’ BABB-type magmatism beginning about 1.8 Ma shows good agreement with the observed Mariana Trough ago. Amagmatic rifting was established after 1.8 Ma. peridotite data, although the critical melting model (a=0.01) also may account for the observed data The Central Graben peridotites could be the residues of low degrees of melting of adepleted MORB-type upper mantle under relatively hydrous condition. Table 9. Mineral/melt partition coefficients used in the models. (Data source: Sobolev and Shimizu 1992) Olivine Orthopyroxene Clinopyroxene Spinel Conclusions Ce 0.0000 0.0006 0.0550 0.0005 Nd 0.0001 0.0090 0.1500 0.0008 The following conclusions are drawn from the data and Sm 0.0007 0.0100 0.2500 0.0009 interpretations presented in this paper: Eu 0.0010 0.0192 0.3200 0.0009 Dy 0.0040 0.0210 0.3500 0.0015 1. The porphyroclastic textures, modal composition, Er 0.0070 0.1000 0.3700 0.0045 and mineral chemistry of Mariana Trough Central Yb 0.0050 0.0444 0.3700 0.0045 Graben peridotites indicate these are residual after Ti 0.0080 0.0805 0.2300 0.1000 Zr 0.0050 0.0225 0.0750 0.0500 melting. 2. Petrologic data indicate that the Central Graben peridotites are ‘‘moderately depleted’’ in a global Table 10. Phase proportions used in the models sense. 3. Small-scale heterogeneity in modal compositions Initial mode Melting mode exists in the upper mantle beneath the Mariana Olivine 0.58 –0.30 Trough Central Graben, although there is no evi- Orthopyroxene 0.26 0.40 dence for pervasive metasomatism or melt–mantle Clinopyroxene 0.13 0.82 interaction. Spinel 0.03 0.08 4. Trace element compositions of clinopyroxenes in Initial mode: Sobolev and Shimizu (1992); melting mode: Kinzler Central Graben peridotites roughly correspond to and Grove (1992) theoretical residual compositions from fractional melting of adepleted MORB-type upper mantle. mechanical extension was more important afterwards in Modeling of open-system melt generation suggests the Central Graben. We infer that mantle flow beneath that 7% melting (F=0.07) of adepleted source the Central Graben is episodic, resulting in varying mantle with 0.5 wt% of interstitial retained melt 16 (a=0.005) at the mass influx rate 0.003 (b=0.003) rifting with relatively ‘‘wet’’ BABB-type magmatism can reproduce the observed REE patterns of Central before 1.8 Ma ago. After the initial magmatic rifting, Graben peridotite clinopyroxenes. amagmatic rifting occurred after 1.8 Ma. 5. Mantle exposures in the Mariana Trough Central Graben may be located within a segment boundary developed in aslow-spreadingridge. The low degree of Acknowledgements We are very grateful to the crew and scientists melting (7%) estimated for Central Graben perido- of the R/V Yokosuka and DSV Shinkai 6500 for their professional tites are consistent with a low magma budget expected work. We thank Kazuhito Ozawa for critical discussions on open- for aslow-spreading ridge. system melting model. We appreciate Nobumichi Shimizu for his 6. We infer that the mantle flow beneath the Central help with the ion microprobe work at Woods Hole Oceanographic Institution. We also thank Henry Dick for fruitful discussion and Graben is episodic, resulting in varying magma supply encouragement. Comments from Kevin Johnson and an anony- rate at spreading segments. Our preferred model for mous referee improved this paper. GMT 3.0 (Wessel and Smith the Central Graben evolution starts with magmatic 1995) was used to map bathymetric data.

Appendix A Results of the simulation (for REEs) discussed in the text Calculated Inﬂuxed Modeled clinopyroxene Accumulated melt initial melt clinopyroxene 1234567810119

(Normalized) CeN 11.22 308.62 0.55 2.92 0.00 0.00 0.01 0.19 0.07 9.92 16.39 23.60 29.86 NdN 12.28 175.42 1.76 8.14 1.17 0.03 0.33 0.60 0.83 11.73 19.28 26.44 33.45 SmN 15.58 149.74 3.59 12.39 4.17 0.54 1.43 1.81 2.91 14.36 22.86 29.72 35.92 EuN 15.74 76.39 4.74 13.42 6.33 1.59 2.69 2.90 4.67 14.81 22.81 28.32 33.10 GdN 15.96 138.43 4.53 13.32 5.85 1.46 2.48 2.88 4.36 15.00 23.28 29.33 34.55 DyN 16.34 127.11 5.47 14.26 7.53 2.38 3.54 3.96 5.80 15.64 23.70 28.93 33.17 ErN 11.81 114.11 5.22 10.84 7.35 3.91 4.54 4.87 6.21 14.12 20.30 23.39 25.67 YbN 14.85 111.60 5.65 13.29 7.94 3.25 4.24 4.59 6.37 15.28 22.65 26.93 30.28 where: 1 equilibrium melting (F=0.15); 2 fractional melting (a=0, melting (F=0.15); 9 fractional melting (a=0, b=0, F=0.05); 10 b=0, F=0.01); 3 fractional melting (a=0, b=0, F=0.05); 4 frac- critical melting (a=0.01, b=0, F=0.1); 11 open-system melting tional melting (a=0, b=0, F=0.1); 5 critical melting (a=0.01, (a=0.005, b=0.003, F=0.07). Inﬂuxed melt composition is calcu- b=0, F=0.1); 6 open-system melting (a=0.01, b=0.01, F=0.1); 7 lated from the clinopyroxene composition of 359R2-2. Normalizing open-system melting (a=0.005, b=0.003, F=0.07); 8 equilibrium value from Anders and Ebihara (1982). Gd values are interpolated

Appendix B Results of the simulation (for Ti and Zr) discussed in the text abF Ti (ppm) Zr (ppm)

Calculated initial clinopyroxene – – – 4,287.67 28.13 Inﬂuxed melt – – – 11,771.44 924.27 Equilibrium melting – – 0.01 3,786.67 19.22 – – 0.05 2,580.55 8.48 – – 0.10 1,845.69 4.99 – – 0.15 1,436.59 3.54 Fractional melting – – 0.20 1,175.95 2.74 – – 0.25 995.35 2.23 0 0 0.01 3,746.44 17.54 0 0 0.05 2,055.42 2.18 0 0 0.10 804.42 0.09 Critical melting 0.01 0 0.01 3,786.66 19.22 0.01 0 0.05 2,284.92 4.98 0.01 0 0.10 1,062.41 0.71 0.01 0 0.15 390.22 0.07 Open-system melting 0.01 0.01 0.01 3,785.03 19.39 0.01 0.01 0.05 2,282.90 5.51 0.01 0.01 0.10 1,069.66 1.40 0.01 0.01 0.15 409.85 0.80 0.01 0.01 0.20 120.90 0.74 Open-system melting 0.005 0.003 0.01 3,776.79 18.94 0.005 0.003 0.05 2,181.78 3.84 0.005 0.003 0.10 941.78 0.54 0.005 0.003 0.15 309.89 0.24

Inﬂuxed melt composition is calculated from the clinopyroxene composition of 359R2-2. 17

Gribble RF, Stern RJ, Newman S, Bloomer SH, O’Hearn T (1998) References Chemical and isotopic composition of lavas from the northern Mariana Trough: implications for magmagenesis in back-arc Anders E, Ebihara M (1982) Solar-system abundances of the ele- basins. J Petrol 39:125–154 ments. Geochim Cosmochim Acta46:2363–2380 Hart SR, Glassley WE, Karig DE (1972) Basalts and sea floor Arai S (1994) Compositional variation of olivine–chromian spinel spreading behind the Mariana Island Arc. Earth Planet Sci Lett in Mg-rich magmas as a guide to their residual spinel perido- 15:12–18 tites. J Volcanol Geotherm Res 59:279–293 Hawkins JW, Melchior JT (1985) Petrology of Mariana Trough Arai S, Matsukage K (1996) Petrology of the gabbro–troctolite– and Lau Basin basalts. J Geophys Res 90:11431–11468 peridotite complex from Hess Deep, Equatorial Pacific: impli- Hawkins JW, Macdougall JD, Volpe AM (1990) Petrology of the cations for mantle–melt interaction within the oceanic litho- axial ridge of the Mariana Trough back-arc spreading center. sphere. In: Me´ vel C, Gillis KM, Allan JF, Meyer PS (eds) Earth Planet Sci Lett 100:226–250 Proceedings of the Ocean Drilling Program, Scientific Results Hussong DM, UyedaS (1981) Tectonic processes andthe history 147, ODP, College Station, Texas, pp 135–155 of the Mariana arc synthesis of the results of Deep Sea Drilling Baker N, Fryer P, Martinez F, Yamazaki T (1996) Rifting history Leg 60. In: Hussong DM, Uyeda S, Blanchet R et al. (eds) of the northern Mariana Trough: SeaMARC II and seismic Initial Reports Deep Sea Drilling Project 60, US Government reflection surveys. J Geophys Res 101:11427–11455 Printing Office, Washington, DC, pp 909–929 Bence AE, Albee AL (1968) Empirical correction factors for the Jaques AJ, Green DH (1980) Anhydrous melting of peridotite electron microanalysis of silicates and oxides. J Geol 76:382– at 0–15 kb pressure and the genesis of tholeiitic basalts. Contrib 403 Mineral Petrol 73:287–310 Bloomer SH, Taylor B, MacLeod CJ, Stern RJ, Fryer P, Hawkins Jaroslow GE, Hirth G, Dick HJB (1996) Abyssal peridotite myl- J, Johnson L (1995) Early arc volcanism and ophiolite problem: onites: implications for grain-size sensitive flow and strain a perspective from drilling in the Western Pacific. In: Taylor B, localization in the oceanic lithosphere. Tectonophysics 256:17– Natland J (eds) Active margins and marginal basins of the 37 Western Pacific. Am Geophys Union Geophys Monogr 88:1–30 Johnson KTM, Dick HJB (1992) Open system melting and tem- Bonatti E, Peyve A, Kepezhinskas P, Kurentsova N, Seyler M, poral and spatial variation of peridotite and basalt at the Skolotnev S, Udintsev G (1992) Upper mantle heterogeneity Atlantis II Fracture Zone. J Geophys Res 97:9219–9241 below the Mid-Atlantic Ridge, 0–15N. J Geophys Res Johnson KTM, Dick HJB, Shimizu N (1990) Melting in the oceanic 97:4461–4476 upper mantle: an ion microprobe study of diopsides in abyssal Bonatti E, Seyler M, Sushevskaya N (1993) A cold suboceanic peridotites. J Geophys Res 95:2661–2678 mantle belt at the Earth’s equator. Science 261:315–320 Johnson KTM, Fisk MR, Naslund HR (1995) Geochemical char- Cannat M, Seyler M (1995) Transform tectonics, metamorphic acteristics of refractory silicate melt inclusions from Leg 140 plagioclase and amphibolitization in ultramafic rocks of the diabases. In: Erzinger J, Becker K, Dick HJB, Stokking LB Vema transform fault (Atlantic Ocean). Earth Planet Sci Lett (eds) Proceedings of the Ocean Drilling Program, Scientific 133:283–298 Results 137/140, ODP, College Station, Texas, pp 131–139 Cannat M, Bideau D, Bougault H (1992) Serpentinized peridotites Karig DE (1971) Origin and development of marginal basins in the and gabbros in the Mid-Atlantic Ridge axial valley at 1537¢N Western Pacific. J Geophys Res 76:2542–2561 and 1652¢N. Earth Planet Sci Lett 109:87–106 Karson JA, Thompson G, Humphris SE, Edmond JM, Bryan WB, Cannat M, Me´ vel C, Maia M, Deplus C, Durand C, Gente P, Ag- Brown JR, Winters AT, Pockalny RA, Casey JF, Campbell rinier P, Belarouchi A, Dubuisson G, Humler E, Reynolds J AC, Klinkhammer G, Palmer MR, Kinzler RJ, Sulanowska (1995) Thin crust, ultramafic exposures, and rugged faulting MM (1987) Along axis variation in seafloor spreading in the patterns at the Mid-Atlantic Ridge (22–24N). Geology 23:49–52 MARK area. Nature 328:681–685 Dick HJB (1989) Abyssal peridotites, very slow spreading ridges Kinzler RJ, Grove TL (1992) Primary magmas of mid-ocean ridge and ocean ridge magmatism. In: Saunders AD, Norry MJ (eds) basalts. 1. Experiments and methods. J Geophys Res 97:6885– Magmatism in the ocean basins. Geol Soc Spec Publ 42:71–105 6907 Dick HJB, Bullen T (1984) Chromian spinel as a petrogenetic Martinez F, Fryer P, Baker NA, Yamazaki T (1995) Evolution of indicator in abyssal and alpine-type peridotites and spatially back-arc rifting: Mariana Trough 20–24N. J Geophys Res associated lavas. Contrib Mineral Petrol 86:54–76 100:3807–3827 Dick HJB, Bryan WB, Thompson G (1981) Low angle faulting and Michael PJ (1988) The concentration, behavior, and storage of steady state emplacement of plutonic rocks at ridge-transform H2O in the suboceanic upper mantle: implications for mantle intersections. EOS Trans Am Geophys Union 62:406 metasomatism. Geochim Cosmochim Acta 52:555–566 Dick HJB, Fisher RL, Bryan WB (1984) Mineralogic variability of Mutter JC, Karson JA (1992) Structural processes at slow- the uppermost mantle along mid-ocean ridges. Earth Planet Sci spreading ridges. Science 257:627–634 Lett 69:88–106 Mysen BO, Kushiro I (1977) Compositional variation of coexisting Elthon D (1992) Chemical trends in abyssal peridotites: refertil- phases with degree of melting of peridotite in the upper mantle. ization of depleted suboceanic mantle. J Geophys Res 97:9015– Am Mineral 62:843–865 9025 Niida K (1997) Mineralogy of MARK peridotites: replacement Fryer P (1995) Geology of the Mariana Trough. In: Taylor B (ed) through magma channeling examined from Hole 920D, MARK Back-arc basins: tectonics and magmatism. Plenum Press, New area. In: Karson JA, Cannat M, Miller DJ, Elthon D (eds) York, pp 237–279 Proceedings of the Ocean Drilling Project, Scientific Results Gaetani GA, Kent AJR, Grove TL, Hutcheon ID, Stolper EM 153, ODP, College Station, Texas, pp 265–275 (1997) Experimental determination of the partitioning of trace Ohara Y, Kasuga S, Ishii T (1996) Peridotites from the Parece Vela elements between peridotite minerals and silicate melts with Rift in the Philippine Sea: upper mantle material exposed in an variable water contents. EOS Trans Am Geophys Union extinct back-arc basin. Proc Jpn Acad Ser B 72:118–123 78:839–840 Ohara Y, Yoshida T, KatoY, Kasuga S (2001) Giant megamul- Gast PW (1968) Trace element fractionation and the origin of lion in the Parece Vela back-arc basin. Mar Geophys Res tholeiitic and alkaline magma types. Geochim Cosmochim Acta 22:47–61 32:1057–1086 Ozawa K, Shimizu N (1995) Open-system melting in the upper Gribble RF, Stern RJ, Bloomer SH, Stuben D, O’Hearn T, New- mantle: constraints from the Hayachine–Miyamori ophiolite, man S (1996) MORB mantle and subduction components northeastern Japan. J Geophys Res 100:22315–22335 interact to generate basalts in the southern Mariana Trough Ross K, Elthon D (1997) Extreme incompatible trace-element back-arc basin. Geochim Cosmochim Acta 60:2153–2166 depletion of diopside in residual mantle from south of the Kane 18

Fracture Zone. In: Karson JA, Cannat M, Miller DJ, Elthon D Trough studied with ‘‘Shinkai 6500’’. Jpn Mar Sci Tech Center (eds) Proceedings of the Ocean Drilling Project, Scientific J Deep SeaRes 13:47–61 Results 153, ODP, College Station, Texas, pp 277–284 Stolper E, Newman S (1994) The role of water in the petrogenesis Sandwell DT, Smith WHF (1997) Marine gravity anomaly from of Mariana Trough magmas. Earth Planet Sci Lett 121:293–325 Geosat and ERS 1 Satellite altimetry. J Geophys Res Taylor B, Klaus A, Brown GR, Moore GF, Okamura Y, Mura- 102:10039–10054 kami F (1991) Structural development of Sumisu Rift, Izu- Seyler M, Bonatti E (1997) Regional-scale melt-rock interaction in Bonin arc. J Geophys Res 96:113–129 lherzolitic mantle in the Romanche Fracture Zone (Atlantic Tucholke B, Lin J, Kleinrock M (1998) Megamullions and mullion Ocean). Earth Planet Sci Lett 146:273–287 structure defining oceanic metamorphic core complexes on the Shaw DM (1970) Trace element fractionation during anatexis. Mid-Atlantic Ridge. J Geophys Res 103:9857–9866 Geochim Cosmochim Acta34:137–143 Volpe AM, Macdougall JD, Hawkins JW (1987) Mariana Trough Shimizu N, Semet MP, Allegre CJ (1978) Geochemical application basalts (MTB): trace element and Sr–Nd isotopic evidence for of quantitative ion-microprobe analysis. Geochim Cosmochim mixing between MORB-like and arc-like melts. Earth Planet Sci Acta42:1321–1334 Lett 82:241–254 Sinton JM, Fryer P (1987) Mariana Trough lavas from 18N: Volpe AM, Macdougall JD, Lugmair GW, Hawkins JW, Lonsdale implications for the origin of back arc basin basalts. J Geophys P (1990) Fine-scale isotopic variation in Mariana Trough bas- Res 92:12782–12802 alts: evidence for heterogeneity and a recycled component in Smith WHF, Sandwell DT (1997) Global seafloor topography from back-arc basin mantle. Earth Planet Sci Lett 100:251–264 satellite altimetry and ship depth soundings. Science 277:1957– Wan W, Yurimoto H (1993) Analysis of rare earth elements in 1962 garnet by SIMS. Annu Rep Geosci Univ Tsukuba 19:87–91 Sobolev AV, Shimizu N (1992) Superdepleted melts and ocean Wessel P, Smith WHF (1995) New version of the Generic Mapping mantle permeability. Doklady Rossiyskoy Akademii Nauk Tools released. EOS Trans Am Geophys Union 76:329 326:354–360 Yamazaki T, Stern RJ (1997) Topography and magnetic vector Sobolev AV, Shimizu N (1993) Ultra-depleted primary melt anomalies in the Mariana Trough. Jpn Mar Sci Tech Center included in an olivine from the Mid-Atlantic Ridge. Nature J Deep SeaRes 13:31–45 363:151–154 Yamazaki T, Murakaki F, Saito E (1993) Mode of seafloor Stern RJ, Lin P, Morris JD, Jackson MC, Fryer P, Bloomer SH, spreading in the northern Mariana Trough. Tectonophysics Ito E (1990) Enriched back-arc basin basalts from the northern 221:207–222 Mariana Trough: implications for the magmatic evolution of Yamazaki T, Okino K, Hasegawa Y, Saitake H, Ito M (1999) back-arc basins. Earth Planet Sci Lett 100:210–225 Geophysical mapping of Mariana Trough and West Philippine Stern RJ, Bloomer SH, Martinez F, Yamazaki T, Harrison TM Basin: a preliminary report of Kairei KR98-12 cruise. Jpn Mar (1996) The composition of back-arc basin lower crust and upper Sci Tech Center J Deep SeaRes 15(II):63–72 mantle in the Mariana Trough: a first report. Island Arc 5:354– Yurimoto H, Yamashita A, Nishida N, Sueno S (1989) Quantita- 372 tive SIMS analysis of GSJ rock reference samples. Geochem Stern RJ, Yamazaki T, Danishwar S, Sun C-H (1997) Back-arc J 23:215–236 basin lower crust and upper mantle in the northern Mariana